Problem description:
Since several weeks we realized that some days, the system clock
of a dual pentium system goes back by 10/15 minutes some days,
meanwhile other days the clock was correct.
The same problem was discovered on a second dual Pentium system
and reported to this forum.
Solution:
According with Clayton Weaver (cgweav@eskimo.com) we had two solutions
to turn around this problem:
1) assuming the hardware clock was uncorrected, use nist package
2) assuming the hardware clock was correct, use hwclock
The hardware clock is stable, this means in 10 days it didn't went back
on both systems.
We have introduced by cron on each hour the command
hwclock --hctosys
and both system clocks are now fine.
Feeling:
The fact the system clock was degrading some days and not other days
is probably due by the fact the days the system clock was corrupted
the system executed heavy disk I/O night background tasks.
Both systems have an internal adaptec scsi controller addressing
4 9GB scsi disks. Then one system is using DPT controller meanwhile
the second is using Mylex eXtremeRAID 1100 for RAID5 storage.
Both systems are using the same network controller (3Com 3C905)
My feeling is that the timer interrupts are delayed by some scsi driver
interrupt handler.
This is a copy of current /proc/interrupts:
0: 16382786 timer
1: 1643 keyboard
2: 0 cascade
8: 3 + rtc
11: 10953720 eth0
12: 1293394 aic7xxx
13: 8687408 + IPI
Hope this will help
Regards
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/